← Cases
TaiwanRetail and e-commerce2025 – ongoing

momo.com

LLM customer service bot built on Azure OpenAI with RAG

momo.com's new AI customer service bot answered correctly more than 90% of the time

WorkingCompany's own claim

Problem

The online retailer needed a customer service system that could answer consistently and cope with traffic peaks such as Singles' Day.

Approach

momo, through its technology unit Fubon Media Technology, built the system with Microsoft Taiwan on Azure OpenAI (GPT models). Retrieval-augmented generation grounds answers in set sources. Complex questions go to human agents. The system went live in July 2025.

Results

above 90%reply correctness of the bot, as reported in a Microsoft Taiwan partnership articleCompany's own claim[1]

Takeaway

momo paired a grounded model with a human fallback for hard questions. Emotion recognition and an AI shopping advisor were plans, not results.

Categories

Business function
Customer service
Innovation type
Process innovation
Evidence
Company's own claim

Sources

Source summaries

[1]TechNews20 Oct 2025

Describes momo's customer service system built with Microsoft Taiwan on Azure OpenAI and RAG, live since July 2025. It gives reply correctness above 90% and says complex questions go to humans. It also lists next steps such as emotion recognition and an AI shopping advisor.

Full details are in the original sources.

Added 2 Nov 2026 · Updated 2 Nov 2026
Open case page ↗